Data Collection

We reviewed the records of infants born during the period January 1, 2010 through December 31, 2014 who were screened for retinopathy of prematurity. Infants with BW of 1500 g or more and who were referred for ROP examination by their attending neonatologist or pediatrician were included in the study. All the babies in the case and control groups were followed according to the same criteria. The criteria for ROP screening for the babies over 1500 g of BW were sepsis and/or requirement of cardiorespiratory support. Infants from diabetic mothers were defined as the case group. Controls were matched 4:1 with cases for GA and year of birth. Neonates with congenital or chromosomal anomalies or major ocular abnormalities in 1 or both eyes and those who died or were lost to follow-up were excluded.

Recommendations of the American Academy of Ophthalmology, the American Academy of Pediatrics, and the American Association for Pediatric Ophthalmology and Strabismus were followed for ROP examination and screening schedule. The examinations were performed using an eyelid speculum, scleral depressor, binocular indirect ophthalmoscope, and 20 diopter lens after pupillary dilation. The “International Classification of ROP Revisited” was used to classify, diagram, and record the retinal findings at the time of examination. Treatment algorithms were based on the recommendations of the “Early Treatment for ROP” and the infants with type 1 ROP were treated with diode laser photocoagulation.

Demographic information and risk factors including BW, GA, singleton or multiple gestations, oxygen therapy with continuous positive airway pressure (CPAP), mechanical ventilation (MV), RDS, hemodynamically significant patent ductus arteriosus (PDA), culture-proven sepsis, necrotizing enterocolitis (NEC, ≥stage 2), intraventricular hemorrhage (IVH, ≥grade 2), bronchopulmonary dysplasia, neonatal glucose irregularity (blood glucose >180 mg/dL and/or <40 mg/dL), assisted reproduction, pre-eclampsia, and maternal diabetes were extracted from the infant’s medical records. Target range of oxygen saturation in our neonatal intensive care unit (NICU) was 85%–94% during the study period.

Maternal diabetes mellitus was recorded as gestational (GDM) or pregestational; however, the results (the rate of ROP and the rate of type 1 ROP) were similar with pregestational (n = 7) and gestational (n = 71) diabetic mothers, and their data were pooled for analyses. The method for diagnosis of gestational diabetes is a glucose challenge test (50 g) as a screening test followed by a complete oral glucose tolerance test (OGTT) (100 g) in case of abnormal screen. Cutoff level for 50 g glucose challenge test was 140 mg/dL. Cutoff levels for 100 g OGTT test were 95 mg/dL for fasting blood glucose, 180 mg/dL for 1 hour, 155 mg/dL for 2 hours, and 140 mg/dL for 3 hours after OGTT. GDM was diagnosed if 2 plasma glucose levels were detected above the cutoff value. Screens were universally performed between 24 and 28 weeks of gestation.

Power Analysis

The sample size needed for the study was computed based on the primary dependent variable of the development of ROP with the binary independent variable of maternal diabetes. The R ² was assumed as 0.50 when the maternal diabetes was regressed on other independent variables; odds ratio was chosen as 20 (under the assumption that the probability of ROP is 0.70 for maternal diabetes [+] infants and 0.10 for maternal diabetes [−] infants according to our pilot study). With these specifications, a sample size of 336 observations achieves 96% power at a 0.05 significance level. Power analysis was done via G*Power 3.

Statistical Analysis

Statistical analyses were performed using SPSS version 16.0 (SPSS Inc, Chicago, Illinois, USA). Frequency (percent) for categorical variables and mean ± standard deviation for metric variables were given as descriptive statistics. The χ ² test was used to compare 2 independent groups for categorical variables, and Student t test was used for metric variables. Phi correlation coefficient was used to assess the degree of association between categorical variables. In order to identify risk factors of outcome variables (the development of ROP and for the development of type 1 ROP), multivariable logistic regression analysis was used. Prior to multivariable logistic regression analysis, the association of each independent variable with the outcome variables, a univariate estimate was performed. The crude and adjusted odds ratio (OR) values and their 95% confidence interval (CI) were given. Although the entry and removal criteria (in terms of P values) were .05 and .10 in multivariable logistic regression analysis, P < .05 was considered as statistically significant in univariate analyses.